Fig. 1.

The proposed joint model based deep learning (J-MoDL) architecture and the training process. Each iteration consists of a CNN block ${\mathcal{D}}_{\Phi }$ and a data-consistency block ${\mathcal{Q}}_{\Theta }$. This architecture facilitates the decoupling of the image priors and the sampling pattern, thus allowing efficient optimization of the parameters Φ and Θ.